GF-1和MODIS影像冬小麦长势监测指标NDVI的对比

作物长势是农情遥感监测的重要内容之一。长期以来,作物长势遥感监测主要基于卫星影像反演的相关植被参数,如归一化植被指数(NDVI,normalized difference vegetation index)、叶面积指数(LAI,leaf area index)等。本文通过对比研究16 m分辨率GF-1卫星影像及250 m分辨率MODIS影像的NDVI与冬小麦综合茎数、株高、叶绿素浓度之间的关系,尝试建立遥感监测作物长势指标与地面实测作物长势指标的定量关系。研究发现GF-1的NDVI与冬小麦综合茎数的相关性最高(R2=0.8961),而与其他指标相关性较弱;MODIS的NDVI指数与冬小麦综合茎数相关性较低(R2=0.4432),对作物长势的遥感监测精度较低。统计MODIS冬小麦像元内GF-1像元的NDVI平均值,并与MODIS的NDVI对比,发现两者之间的相关性较低(R2=0.3944);在消除MODIS与GF-1影像传感器光谱响应函数差异及NDVI尺度效应后,MODIS影像的冬小麦作物长势遥感监测精度得到一定提高(R2=0.4633)。对MODIS像元内GF-1 NDVI标准差排序发现,MODIS像元内冬小麦长势一致性越高,MODIS的长势遥感监测精度越高。GF-1和MODIS影像NDVI长势监测主要代表地面冬小麦综合茎数,且卫星影像分辨率越高,NDVI值越能反映实际的作物长势。MODIS像元内冬小麦长势一致性越高,基于NDVI的MODIS与GF-1数据冬小麦长势监测结果越一致。从区域长势监测角度来看,尽管MODIS与GF-1数据的监测结果趋势较为一致,并且通过光谱、尺度归一化能够进一步提高监测结果的一致性,但MODIS NDVI长势监测总体精度较低,为满足作物长势精细化监测的业务需要,应逐步使用高分辨率的遥感数据替代中低分辨率遥感数据进行作物长势遥感监测,并将其作为长势监测业务化运行的研究重点。

Comparison of Growth Monitoring Index NDVI between GF-1 and MODIS Images in Winter Wheat

Crop growth status is one of the major contents of agriculture remote sensing monitoring. For a long time, crop growth status remote sensing monitoring has been mainly based on related vegetation parameters inverted from satellite images, such as normalized difference vegetation index (NDVI), leaf area index (LAI). In this study, we made an attempt to set up a quantitative relation between remote sensing monitored crop growth indices and ground observation indices of actual crop growth, by comparative study of NDVI based on the GF-1 satellite image with resolution of 16 m and the MODIS image with the resolution of 250 m and comprehensive stalk number (CSN), plant height (PH), and leaf chlorophyll concentration (LCC) of winter wheat. GF-1 NDVI had the highest correlation with CSN (R ² = 0.8961), while other indices had relatively low correlation. In contrast, MODIS NDVI had relatively lower correlation with CSN of winter wheat (R ² = 0.4432), and the accuracy in crop growth remote sensing monitoring was relatively low. The average NDVI value of GF-1 pixels within MODIS winter wheat pixels was measured and found to be weakly correlated with MODIS NDVI (R ² = 0.3944). However, the wheat growth monitoring accuracy by MODIS images was improved (R ² = 0.4633) after eliminating the sensor spectral response function difference and scale effects. By ranking the standard deviations of GF-1 NDVI within the MODIS pixels, we found that higher winter wheat growth consistency within MODIS pixels resulted in higher monitoring accuracy of MODIS remote sensing. The NDVI from GF-1 or MODIS image mainly represented CSN, and the NDVI value from higher solution image had a better reflection of actual winter wheat growth. When the winter wheat growth consistency within MODIS pixels was high, the difference of NDVI between MODIS and GF-1 data was small. From the perspective of long term regional growth monitoring, although spectrum and scale normalization can further improve the consistency of monitoring result, the overall accuracy of MODIS NDVI growth monitoring is relatively low. In order to meet the demand of precise monitoring on crop growth, it is necessary to use high-resolution remote sensing data to replace middle- or low-resolution data for crop growth remote sensing monitoring, and take it as the research focus of the business operation of crop growth monitoring.